Priority Assignment in Real-Time Active Databases 1. Bhaskar Purimetla, Rajendran M. Sivasankaran, John A. Stankovic, Krithi Ramamritham & Don Towsley

Size: px
Start display at page:

Download "Priority Assignment in Real-Time Active Databases 1. Bhaskar Purimetla, Rajendran M. Sivasankaran, John A. Stankovic, Krithi Ramamritham & Don Towsley"

Transcription

1 Priority Assignment in Real-Time Active Databases 1 Bhaskar Purimetla, Rajendran M. Sivasankaran, John A. Stankovic, Krithi Ramamritham & Don Towsley Department of Computer Science University of Massachusetts, Amherst, MA Abstract Active databases and real-time databases have been important areas of research in the recent past. It has been recognized that many benets can be gained by integrating active and real-time database technologies. However, there has not been much work done in the area of transaction processing in active real-time databases. This paper deals with an important aspect of transaction processing in active real-time databases, namely the problem of assigning priorities to transactions. In these systems time-constrained transactions trigger other transactions during their execution. We present various policies for assigning priorities to parent, immediate and deferred transactions executing on a multiprocessor system and then evaluate the policies using simulation. The simulator has been validated by comparing to three sets of published results. Our new results demonstrate that dynamically changing the priorities of transactions depending on their behavior (triggering rules), yields a substantial improvement in the number of triggering transactions that meet their deadline. 1 Introduction Traditionally, in soft real-time transaction processing systems, a transaction is considered a monolithic unit of work with a given deadline. In these systems, priorities are assigned to these transactions and the transactions are scheduled based on their priorities. The priority assignment usually takes into account the deadline of the transactions, because the underlying assumption is that the deadline reects the urgency of completing the transaction. Scheduling policies such as Earliest Deadline First (EDF) and Least Slack First (LSF) are examples of priority-cognizant policies. The performance implications of priority assignment policies have been studied in detail in soft real-time transaction processing systems in [1, 6]. In such systems, the priority assignment policies as well as conict resolution policies must be time cognizant. In this paper, our goal is to study and evaluate priority assignment policies in a real-time active database. A real-time active database is a database system where transactions have timing constraints such as deadlines, where transactions may trigger other transactions, and where data may become invalid with the passage of time. There are many applications such as cooperative distributed navigation systems and intelligent network services where real-time active database technology is extremely useful [12, 14]. Before explaining the problem we are addressing in detail, we will give a brief introduction to active databases. The building block of an active database system is the Event-Condition-Action 1 This work was supported in part by NSF under grants IRI and IRI

2 (ECA) rule. The semantics of the ECA rule is that if the specied event (E) occurs and if the condition (C) is true then the specied action (A) is to be executed. Some examples of events are begin 2, commit/abort of a transaction, accessing a data item, or reaching a specic point in time. A condition is usually a predicate on the database state. An action is the transaction that is executed in reaction to the specic situation which is the combination of events and the conditions. The transaction that res the rules is called the triggering transaction and the action that is executed because of the rule ring is called the triggered transaction. We refer to the transactions that trigger other transactions as active transactions or parent transactions in this paper. An active transaction has a set of triggered transactions that are either executed as part of the active transaction or separately { depending on the type of the coupling mode between the parent and the triggered transactions [4]. There are three types of coupling modes. They are immediate, deferred and independent and the transactions triggered in those modes are referred to as immediate, deferred and independent transactions, respectively. The immediate and deferred transactions are executed as part of the parent transaction whereas the independent transactions are executed independently. An immediate transaction is executed as soon as it is triggered and the parent transaction is suspended until the immediate transaction is completed. A deferred transaction is executed after the parent nishes execution, but before it commits. The immediate and deferred transactions are committed only if the parent commits. Since the immediate and deferred transactions are part of the parent transaction, we also refer to them as subtransactions. Due to the rule rings, an active transaction generates additional work dynamically. In a real-time context, for the time-cognizant scheduling policies to perform well, they should take into account the dynamic work that is being generated. This aspect of the problem makes this scheduling problem dierent from the classical hard real-time scheduling where execution times are assumed to be known in advance [8, 9, 15]. The priority driven nature of real-time transaction processing brings about the question of assigning priorities to the parent and all triggered actions [12, 14]. We believe that the priority assignment strategy has a signicant impact on the performance of the system as triggered actions will contend with ongoing transactions for resources. In this paper, we address the problem of assigning priorities to triggering and the triggered transactions. We discuss three policies for assigning priorities to immediate and deferred transactions, given the triggering and triggered transactions' characteristics. To illustrate the problem, we provide an example of the structure of a complex active transaction executing on a uniprocessor. We use gure 1 to illustrate our example. Since we use gure 1 later in our description of priority assignment protocols, it is more complicated than necessary for the problem illustration. Figure 1 shows the life of a complex active transaction T that triggers transactions in all three modes. Transaction T arrives at time t1 (at(t )) with t10 as its deadline (dl(t )) and is started at time t2 (st(t )). It triggers a deferred transaction dt1 at time t 3, and another deferred transaction 2 The begin event may, in turn, have been caused by some external environment event such as an obstacle identied by a sensor [12, 14]. 2

3 ind1 T dt1 dt2 it1 dt3 it2 dt1 dt2 dt3 t1 t2 t3 t4 t5 t6 t7 t8 t9 t10 at(t) st(t) dl(t) Figure 1: Life of a Complex Active Transaction dt2 at time t4. It triggers an independent transaction ind1 at t5, an immediate transaction it1 at t6, another deferred transaction dt3 at t7 and an immediate transaction it2 at t8. Figure 1 shows that the transactions it1 and it2 execute immediately while T is suspended. Finally, the gure shows that once the parent T completes, the deferred transactions dt1, dt2 and dt3 execute. The problem we address in this paper is the problem of assigning priorities to transactions it1, it2, dt1, dt2 and dt3, and the problem of dynamically reassigning the priority of the parent transaction T during its lifetime. It should be noted that the deferred transactions can execute simultaneously if operating on a multiprocessor or a distributed system. We assign the priorities to triggered transactions when they start execution and they are not modied later. Our main contributions are : developing priority assignment policies that take into account the dynamic work generated by active transactions; evaluating the priority assignment policies against a baseline policy using a real-time active database simulator in two settings, i.e., in a real-time task setting and a main memory database setting; demonstrating that for mixed workloads with triggering and non-triggering transactions, priority assignment policies that take into account the dynamic work generated reduce the deadline miss ratio of the triggering transactions signicantly at the cost of a very small increase in the deadline miss ratio of non-triggering transactions compared to the baseline policy; and conducting experiments to identify the relative bias the policies show for the triggering transactions, thereby enabling an implementor to select from various policies depending on the relative importance of the triggering and non-triggering transactions in the system. We discuss related work in Section 2. In Section 3, we explain our transaction model, system model and describe the various attributes of a transaction and other related terms. Section 4 gives a detailed explanation of the priority assignment policies. We discuss the experiments and results 3

4 in Section 5 in detail. We conclude our paper in Section 6 with a summary of our main results and discuss future work. 2 Related Work Over the past few years active databases and real-time databases have become important areas of research. There have been both theoretical and experimental studies in active databases [2], [3], [11], [4]. Experimental work on active databases in [2] is done in a non real-time setting. Most of the research done on active components in object oriented databases and real-time databases has concentrated on the specication of Event-Condition-Action (ECA) rules [5]. Experimental work done on real-time transaction processing systems [1, 6] has not considered active workloads. Experimental studies reported in [1] are very comprehensive and cover most aspects of real-time transaction processing, but have not considered active workloads and have not addressed the problem of subtransaction priority assignment. Experiments in [2] show the impact of transaction boundaries and data sharing on performance of active databases. We address a dierent problem, that of assigning priorities to triggered transactions in real-time active databases. In [13] the relationship between real-time databases and active databases is discussed briey. Lack of active pursuit of timely processing of actions to do real-time processing was identied as a missing ingredient in active databases. In short, past studies on real-time transaction processing have not dealt with active workloads and studies on active databases have not dealt with real-time transactions. Simulation studies have been conducted to study the problem of assigning deadlines to subtasks in real-time systems [7]. The problem of assigning deadlines to the parallel and serial subtasks of complex distributed tasks is addressed in [7]. The structure of the complex tasks is assumed to be known in advance. Also the experimental system considered in [7] is not a transaction processing system and does not have an active component. The system we consider for our experiments is an real-time active database with unpredictable data accesses and rule rings. In contrast with [7], in our study the structure of complex transactions is not known in advance. 3 The Model In this section we describe the transaction model and the system model used to study the priority assignment policies in a real-time active database. We also dene some key attributes of a transaction and other related terms. 3.1 Transaction Model A transaction in our system is a series of method executions on objects. We consider two types of transactions in the system: Triggering (class T) and NonTriggering (class NT). A nontriggering transaction is a simple transaction which does not cause any rules to re. A triggering transaction, 4

5 on the other hand, can trigger subtransactions upon the occurrence of an event. All subtransactions are simple transactions. An event in our model is one of the following types of events: Transaction Event, Object Event, or Temporal Event. An object event occurs, whenever the transaction executes a method on an object instance, a transaction event occurs during begin, commit or abort of a transaction and a temporal event occurs when a particular point in time is reached. The subtransaction that is triggered can be executed in one of the three modes: immediate, deferred or independent. 3.2 Attributes of a Transaction In order to understand the details of priority assignment policies, we dene the following attributes of a transaction T, and other terms that we require to explain our priority assignment policies. at(t ) : The arrival time of T st(t ) : The start time of T dl(t ) : The absolute deadline of T vdl t (T ) : The virtual deadline of T at time t (this is used for scheduling) avglen(t ) : Average length of T (average number of method invocations) lenlef t(t ) : Length Left (remaining number of method invocations based on avglen(t) assumption) slack t (T ) : The slack of T at time t wslack t (T ) : The predicted worst case slack of T at time t eet(t ) : The estimated execution time of T ect t (T ) : The estimated completion time of T at time t eetl t (T ) : The estimated execution time left for T at time t wcet t (T ) : The predicted worst case execution time of T at time t ndef t (T ) : The number of deferred transactions triggered by T until time t nimm t (T ) : The number of immediate transactions triggered by T until time t eetdef t (T ) : The sum of estimated execution time of all deferred transactions triggered until time t probt rig obj : Probability of an object event triggering a subtransaction probt rig trans : Probability of a transaction event triggering a subtransaction The vdl t (T ), the virtual deadline of transaction t at time T is used to assign the transaction's priority by the EDF scheduler. To increase the priority of a transaction we assign a virtual deadline which is earlier than the absolute deadline dl(t ). Thus, it is not necessary to abort a transaction when its virtual deadline expires. eet(t ), estimated execution time information, is used by some priority assignment policies. eetl t (T ), is the dierence between the estimated execution time and the time taken to execute the transaction so far. The wcet t (T ), includes the execution time left at time t as well as the worst case time to execute the subtransactions which the transaction T can trigger during its remaining lifetime. The predicted worst case slack wslack t (T ) is obtained by 5

6 Source DBManager Transactions Object Access Request Transaction Manager Transaction Events Resource Request Rule Manager Resource Manager Object Events Resource Request Object Manager Figure 2: Simulator Architecture subtracting the worst case execution time from the absolute deadline. The calculation of worst case execution time uses the value of probt rig obj and probt rig trans, which are the probabilities that a subtransaction is triggered by an object event or a transaction event of the transaction, respectively. This information is used by some of the priority assignment policies. We believe that by analyzing the characteristics of an application one might be able to obtain information like execution time of transactions, the kind of actions they trigger and the probability of triggering. In our experiments, we assume that this information is known accurately. In future experimentation, we plan to study the implication of errors in this information, on the performance of the priority assignment policies. The following equations show the relationship between the some of the terms dened above. ect t (T ) = clock + eetl t (T ) slack t (T ) = dl(t ) - ect t (T ) where clock is the current clock value. 3.3 Simulation Model Our performance model of an active real-time active database was implemented using the DeNet Simulation Language [10]. The simulator is made up of ve active components : Source, Transaction Manager, Object Manager, Resource Manager, Rule Manager, and a passive component DB Manager. Figure 2 illustrates the architecture of the simulator. Following is a detailed description of the modules: DB Manager This is the passive module that models the data. The data is modeled as having a certain 6

7 number of object classes and each object class having a certain number of instances. Each object class has a certain number of methods dened which are used to access the object. Each object instance in the database is mapped to a page or number of pages in the secondary storage. Source The source (transaction generator) generates the incoming transactions into the system. One can view the source as the application or the environment in which the Real-Time Active Database is used. It generates transactions with timing constraints with a specied arrival distribution. It can generate both periodic and aperiodic transaction streams and three types of real-time transactions namely hard, rm and soft. One of our future goals is to study the transaction characteristics of specic applications [12, 14] and design the generator to model it as closely as possible. Transaction Manager Transaction Manager is responsible for scheduling and execution of transactions it receives from the source. It executes the submitted transactions by requesting the Object Manager to execute the specic methods on specic objects. It handles the various transaction events: begin, commit and abort. It informs the Rule Manager of the transaction events. Object Manager Object Manager executes the methods of the objects in the system as requested by the Transaction Manager. It is also responsible for concurrency control. If the request can be executed it executes and sends the success message back to the Transaction Manager. If the request cannot be satised it either sends a blocked or abort message back to the Transaction Manager. It also informs the Rule Manager of the various object events that occur. Conict resolution in this module is priority based where the lower priority conicting transaction waits or gets aborted depending on whether it is the requester or holder of locks. Rule Manager The Rule Manager models the active workload in the system. The rule rings are modeled probabilistically, i.e., a rule is red with a certain probability. The Rule Manager checks to see if any rules are triggered whenever it gets an event notice from the Transaction Manager or the Object Manager. It models the condition evaluation, and, nally it generates the transactions corresponding to the actions of the rules triggered if their conditions are satised and submits them to the Transaction Manager. Resource Manager The Resource Manager simulates the CPUs and disks and the main memory buer. The Object Manager requests the Resource Manager for the necessary pages or for CPU time to execute the methods. The Transaction Manager requests the Resource Manager the CPU time and buer 7

8 space to load transactions. The CPU, disk, and memory resource scheduling is priority driven. Our resource model is a multiprocessor, multi disk, shared memory system. The incoming resource requests are queued in a common CPU queue or a common disk queue depending on the kind of request. The scheduling and conict resolution decisions made in dierent modules of the simulator are independent of each other. No global scheduling decisions are made. For instance, the scheduling decision made in the Resource Manager is independent of the one made in the Transaction Manager. There are three types of overload management policies available in the simulator. In the all eligible policy all the transactions are run to completion. In the not tardy policy, a transaction is aborted as soon as its deadline expires. This corresponds to rm real-time transactions. This policy assumes that nishing a transaction after its deadline expires doesn't impart any value to the system. In feasible deadline policy, we check the feasibility of a transaction, i.e., if it can nish by its deadline based on the estimated execution time and abort it if it cannot nish by its deadline. 4 Priority Assignment for Triggered Transaction In real-time active databases, a transaction with time constraints can trigger subtransactions. Traditionally, priority driven scheduling has been used in real-time systems. So the problem arises as to how to assign a priority to a subtransaction given the priority of the parent transaction. Another problem is that as a transaction triggers subtransactions either in immediate or deferred modes, the amount of work to be done on behalf of the transaction before it commits also increases. So the transaction is less likely to complete successfully compared to another transaction with the same deadline which doesn't trigger any subtransactions, since it faces more resource contention during its lifetime. In this section, we describe some policies for dynamically assigning priorities to the parent and Immediate and Deferred subtransactions in an active real-time database system. In all cases, the priorities are assigned to the triggered transactions when they start execution and priorities are not changed subsequently during their execution. The priority of the triggering transactions may change dynamically, depending on the policy. 4.1 Priority Assignment Policies for Immediate Subtransactions Three priority assignment policies are studied for immediate subtransactions. The rst two assignment policies assume that the scheduling discipline is Earliest Deadline First (EDF) and the third policy assumes that the scheduling discipline is Least Slack First(LSF). We use the active transaction T in gure 1 to explain our policies and discuss the priority assignment policies for the immediate transaction it1. We request the the reader to refer to section 3.2 to understand the semantics of the terms used in the following description. 8

9 1. PD: Assign the same deadline as the parent deadline. The priority of the parent transaction which is based on its deadline does not change with the triggering of subtransactions. This is a very simple baseline algorithm. All the actions done on behalf of a transaction get the same priority as the transaction itself at any point during its lifetime. For example, at the triggering point, i.e., at time t6 dl(it1) = dl(t ) [1] vdl t6 (it1) = dl(t ) [2] 2. DIV: Equally divide the parent's eective slack among all the immediate and deferred subtransactions triggered until that point. The parent's deadline is also adjusted dynamically to reect the work that has been triggered dynamically. The sum of estimated execution time of deferred transactions that have been triggered until that point is subtracted from the parents slack to determine the eective slack. vdl t6 (it1) = ect t6 (it1) + slack t6(t )?(eet t6 (it1)+eet t6 (dt1)+eet t6 (dt2)) ndef t6 (T )+nimm t6 (T ) [3] dl(it1) = dl(t ) [4] vdl t6 (T ) = vdl t6 (T ) - eet(it1) [5] The main idea behind this policy, is that of giving higher priority to class T transactions, which have more work to do before completion. This will increase the probability of the class T transactions meeting their deadlines. This policy only uses the estimates of execution times of subtransactions that have already been triggered. It does not use any knowledge about future triggering of transactions. 3. SL: Adjust the worst case slack (wslack t (T )) of the parent at each potential triggering point. In this policy, the transaction with the smallest slack has the highest priority. The initial value of slack is assigned based on the predictions about the total execution time for a transaction and its subtransactions as indicated by probt rig obj. The slack is then adjusted at each object or transaction event based on whether the parent transaction triggers a subtransaction or not. The triggered transactions are assigned the same slack as the parent, i.e., they are executed at the same priority. We assume that the transaction events do not trigger any subtransactions. This assumption holds for the rest of the paper and is explained later in more detail. Initially slack is set as follows: wslack t6 (T ) = dl(t ) - clock - eet(t ) - (avglen(t ) * probt rig obj * eet(st)) [6] where st is a triggered subtransaction. If a subtransaction, say it1 is triggered the slack is adjusted as follows: 3 3 Note that we update slack only at object events. 9

10 wslack t6 (T ) = dl(t ) - clock - eetl t6 (T ) - eet(it1) - (lenleft(t ) * probt rig obj * eet(st)) [7] wslack t6 (it1) = wslack t6 (T ) [8] If a subtransaction is not triggered then the slack adjustment is as follows : wslack t6 (T ) = dl(t ) - clock - eetl t6 (T ) - (lenleft(t ) * probt rig obj * eet(st)) [9] For a given transaction T, wslack(t ) is used to calculate its priority during scheduling. The deadlines are assigned as follows: vdl t6 (it1) = dl(t ) [10] dl(it1) = dl(t ) [11] 4.2 Deadline Assignment Policies for Deferred Subtransactions The priority assignment for deferred transactions is very similar to that of immediate transactions. There are three policies, PD, DIV and SL for deferred transactions but in the case of DIV the denominator of equation [3] considers only the deferred transactions that are yet to be executed, because all the immediate transactions would have already nished execution. The following equations illustrate the deadlines/slack assignments in the case of deferred transactions. Once again, we use the active transaction T in gure 1 in our explanation and discuss priority assignment policies for deferred transaction dt1. 1. PD Protocol: 2. DIV Protocol: 3. SL Protocol: dl(dt1) = dl(t ) [12] vdl t9 (dt1) = dl(t ) [13] vdl t9 (dt1) = ect t9 (dt1) + slack t9(t )?(eet t9 (dt1)+eet t9 (dt2)+eet t9 (dt3)) ndef t9 (T ) [14] dl(dt1) = dl(t ) [15] wslack t9 (T ) = dl(t ) - clock - eetdef t9 (T ) [16] wslack t9 (dt1) = wslack t9 (T ) [17] vdl t9 (dt1) = dl(t ) [18] dl(dt1) = dl(t ) [19] 10

11 5 Experimentation and Results We begin this section with a brief description of the validation of our simulator. We then discuss the experimental setup along with the assumptions and decisions made in our experiments. We also present a table of important parameters and their values. Finally, we describe each set of experiments and an analysis of the results. In the experiments, 95% condence intervals have been obtained whose widths are less than 2:5% of the point estimate for the Missed Deadline Percentage (MDP). 5.1 Validation of the Simulator Experiments were conducted to validate the simulator. This was accomplished in three steps. 7 active-validation 6 Response Time IMM/[2] IMM IND/[2] IND Arrival Rate Figure 3: We validated the active part of the simulator against the results in [2]. The results are illustrated in the gure 3. We mapped their model onto ours as closely as possible. We were not able to obtain the exact results because of dierences in the two models. The original work modeled the buer explicitly, but we model our buer using a parameter hitratio, which is the probability that a page is resident in the buer. Hitratio is set to 0.9, for all the experiments in this validation. The explanation of parameters and the experiments can be found in [2]. In gure 3 solid lines (marked as /[2]) represent the original results and the dotted lines represent our simulation results. From gure 3, we can see that for Independent coupling mode (no coupling) and the Immediate coupling mode (strict coupling), our results are within 10 percent of the original results. We validated the real-time part by trying to duplicate the results in [1]. In the NT (not tardy) overload management policy, a transaction is aborted as soon as it becomes tardy. In the AE (all eligible) policy, a transaction is run until it nishes. The results are illustrated in gure 11

12 4. Again our results were very similar to previously published results. The slight performance improvement obtained by our policies in the not tardy case, can be explained by the fact that checking for tardiness is done more often in our model. Finally, to validate the combined model, we conducted experiments as in [7] (No graphs are plotted for these experiments). There was an inherent dierence in the two models because the one in [7] is a distributed system and ours is a single site multiprocessor system. So their system has multiple servers with a queue for each server whereas ours has multiple servers with a single queue. We experimented with UD and DIV-1 policies mentioned in [7] and our missed deadline percentages were lower by no more than 5-10% which can be attributed to the above dierence and other subtle dierences in parameters. 100 realtime-validation Missed Deadline % AE/FCFS/[1] AE/FCFS NT/FCFS/[1] NT/FCFS Arrival Rate Figure 4: In our validations we did not perform the complete set of experiments that are in [2], [1] or [7], but just certain baseline experiments. 5.2 Experimental Setup In our experiments We address only rm real-time systems where the value of the transaction drops to zero once the deadline expires. We abort the transaction that misses its deadline along with all its related subtransactions. This corresponds to the not tardy case. We do priority-driven preemptive scheduling. We consider active real-time databases that trigger only immediate and deferred transactions. 12

13 We disable cascading rule rings, that is triggered transactions do not trigger further transactions. Transaction events do not trigger rules. We make this assumption to simplify the experimentation as the only eect of transaction events triggering rules in our study, would be to increase the number of subtransactions triggered. This eect can be achieved by increasing the probability that an object event can trigger a rule. There are three types of class T transactions: IMM which trigger only immediate subtransactions, DEF that trigger only deferred transactions, and IMMDEF that trigger both immediate and deferred subtransactions. We normalize the slacks for the three types of class T transactions taking into account the fact that the deferred transactions can be executed in parallel and the immediate transactions are executed in sequence. Therefore, transactions that trigger only immediate subtransactions will get more slack than those that trigger only deferred subtransactions. The Missed Deadline Percentage (MDP) is the performance metric we use. This is the traditional metric used to compare performance in rm real-time systems. We experiment with two types of data access characteristics of transactions. a main memory based system with no data conicts. The rst is This is essentially a task model as opposed to transaction model. Second is a main memory database, i.e., main memory based system with data conicts. We have no explicit buer modeling, but a parameter hitratio that probabilistically says if a page is in memory, which is set to 1.0 to simulate main memory database system. Similarly, there is no explicit mapping of objects to specic pages, but every object access is converted in to a corresponding number of page accesses. We use a parameter load in our experiments which is very similar to the one in [7]. In order to dene load we specify the arrival rates and service rates of class T and NT transactions. The arrivals of class T and class NT transactions are generated according to Poisson processes with mean interarrival time of 1= T and 1= NT time units, respectively. The arrival rates are calculated using the following two equations, where all other quantities except the arrival rates are known to us. In the rst equation, we dene the load to be the ratio of work generated to the total processing capacity of the system. 1= T and 1= NT are the average total execution time of class T and NT transactions respectively, and N CP U is the number of CPUs in the system. In the second equation, frac T is the fraction of load that is contributed by the class T transactions. load = T T + NT NT N CP U [] frac T = T T T T + NT NT [21] 13

14 Table 1 shows parameter settings for our baseline experiments. The deadline of a transaction T is set using the following formula: dl(t ) = at(t ) + (1 + slack) eet(t ) where slack is randomly selected from a specied range. Mode Parameter Setting ALL N CP U 6 Number of Object Instances 00 Overload Management Policy Not Tardy Avg. Length of class NT Transactions 5 slack of class NT transactions Avg. Length of class T Transactions 6 Avg. Length of Subtransactions 5 frac T 0.15 Probability of triggering by object event 0.8 Probability of triggering by transaction event 0.0 IMM slack % of Immediate Subtransactions 100 % of Deferred Subtransactions 0 DEF slack % of Immediate Subtransactions 0 % of Deferred Subtransactions 100 IMMDEF slack % of Immediate Subtransactions 50 % of Deferred Subtransactions 50 Table 1 : Baseline setting 5.3 Real-time tasks In the rst set of experiments we deal with real-time active tasks executing in a multiprocessor environment. The purpose of these experiments is to isolate and study the eect of scheduling on performance. We simulate a main memory database system where there are no data conicts. We have a parameter datacontention which, if set to 1.0, makes all the methods of an object class compatible with each other and when set to 0.0 makes all the methods incompatible with each other. So in these experiments, we limited the number of object classes to 1 and the number of methods to 1. Depending on the datacontention the method is either compatible with itself or not. All the object instances in the database belong to this single class. We experiment with all three types of class T transactions, i.e., IMM, DEF and IMMDEF. The performance of the transactions belonging to both 14

15 classes is presented in gures 5, 6 and 7, respectively. Figure 5 deals with the case where all the subtransactions are triggered in immediate mode (IMM). We can see from the graph that compared to PD both the DIV and SL policies decrease the MDP of T class transactions signicantly at higher loads by as much as 10 to 12 percent, at the cost of a small increase of around 3 percent in the MDP of NT class transactions. The protocol DIV reduces the MDP of T class by a greater amount than SL, accompanied by a smaller increase in the MDP of NT class. So clearly DIV performs better than SL in the case of immediate transactions. In gure 6, we present the results for the case where all the subtransactions triggered are executed in deferred mode (DEF). Again, SL and DIV policies perform better than the PD protocol for class T transactions. In this case SL and DIV reverse their roles, SL outperforming DIV substantially at higher loads for T class. However, in the case of NT class DIV performs better than SL. For the T class SL reduces the MDP by 30 percent at high loads compared to DIV with a slight increase in the MDP of NT class. Essentially SL gives higher preference to T class transactions over NT class transactions than DIV. This change of order from the previous result can be explained by the fact that deferred subtransactions can be executed in parallel on a multiprocessor system whereas the immediate subtransactions are executed sequentially. So SL gives a very high preference to T class transactions when deferred subtransactions are present. This explains the fact that SL keeps the MDP of T class transactions nearly constant while the MDP of NT class increases. It also explains the fact that SL gives lower MDP for T class than the NT class. The results for the case where the subtransactions triggered execute either in deferred or immediate mode with equal probability (IMMDEF), are illustrated in gure 7. The performance trend is similar to that of the deferred only case, but the dierence is lower than the DEF case because of the presence of immediate subtransactions. In the above results we observe that while the MDP of T class transactions is reduced by the SL and DIV policies, the MDP of the NT class transactions increases. This is desirable, where T class transactions are more valuable than NT class transactions. To see the performance of the policies if transactions of both classes have the same value, we evaluated the combined MDP of all the transactions. Again 85 percent of the workload comes from NT class transactions. The results are illustrated in gures 8,9 and 10. As we can see PD performs best if we give equal value to both T and NT class transactions. DIV is the next best and the SL is the worst. This is because DIV and SL are biased toward T class transactions. Also the CPU time required by one T class transaction is much higher. Though the MDP of T class is reduced, since NT class transactions constitute the majority, the overall MDP increases in the case of DIV and SL. For example, in gure 8, SL shows the worst performance and DIV is comparable to PD except at very high loads. So DIV protocol performs comparably with the PD protocol even when transactions of both classes have equal value. In DEF and IMMDEF case clearly PD provides the lowest MDP over DIV and SL policies in that order. 15

16 5.4 Analysis of DIV and SL Policies We observed in the previous experiments that DIV and SL give preference to T class transactions over the NT class. In this section we will study what happens when we design algorithms which range between PD and DIV and similarly between PD and SL. So we introduced a parameter which controls the priority assignment in DIV and SL policies as follows. We call these parameterized policies ALPHA-DIV and ALPHA-SL. ALPHA-DIV: priority =? deadline DIV + (1? )? deadline parent deadline DIV is the deadline assigned to the transaction by the DIV protocol. deadline parent is the deadline of the parent transaction. ALPHA-SL : priority =? estslack SL + (1? )? deadline parent estslack SL is the slack assigned to the transaction by the SL protocol. A lower priority value indicates higher importance. So when is zero both the policies reduce to PD. When is one they are DIV and SL policies respectively. For this set of experiments the load was kept constant at We studied the performance of the ALPHA-DIV and ALPHA-SL policies as is changed from zero to one. The results are plotted in gures 11,12 and 13. On X and Y axes we plot the MDP of NT and T class transactions, respectively. The points on the curves, correspond to values varied from 0 to 1 with an increment of 0.2. We also plotted some points with intermediate values of to clearly distinguish between the dierent policies. We observe that ALPHA-DIV works better than ALPHA-SL throughout the range for the immediate only case. It reduces the MDP of T class transactions by 22 percent with a little increase of only 1 percent in the MDP of NT class, compared to the ALPHA-SL protocol which achieves the same eect at the cost of a 3.5 percent increase. So the ALPHA-DIV protocol performs better than the ALPHA-SL protocol in this case. But in the deferred only transactions case the ALPHA-SL protocol reduces the MDP of T class to a very negligible value with a rise in the MDP of NT class. ALPHA-DIV is not able to reduce the MDP of T class transactions that signicantly. So if T class transactions have a higher value than NT class transactions, then ALPHA-SL is the protocol of choice in the deferred only case. ALPHA- DIV marginally outperforms ALPHA-SL at lower values. In the case where subtransactions can be deferred or immediate with equal probability, the trend is the same as the deferred only case. The results are illustrated in gure 13. So ALPHA-SL gives more exibility to achieve a higher reduction in the MDP of T class with an increase in the MDP of the NT class than the ALPHA-DIV policy. 5.5 Main Memory Database We now extend the performance study to consider priority assignment policies in the presence of data contention. The concurrency control algorithm we use is a modied wound-wait algorithm. All the subtransactions of a transaction (deferred and immediate modes) are considered part of the transaction by the concurrency control mechanism. They share the locks. Similarly, two subtransactions 16

17 of the same parent transaction share the locks. All the subtransactions and the parent transaction release the locks at their commit time which is after the parent transaction and the deferred subtransactions have nished. Since dierent subtransactions may have dierent priorities, deadlocks might occur in spite of using the wound-wait protocol. So we use a deadlock prevention mechanism, which checks for deadlocks whenever a transaction waits for another transaction and aborts the transaction that causes the deadlock. It has been shown in previous studies that the choice of the transaction to be aborted to resolve a deadlock does not have a signicant impact on the performance [6]. For these experiments we set the parameter datacontention to 0.0, making any two method accesses incompatible. This is equivalent to locking the whole object exclusively for each operation. We keep the rest of the parameter values the same as in previous experiments. The results are illustrated in gures 14,15 and 16 for immediate only, deferred only and both immediate and deferred cases, respectively. The relative ordering of the performance of the policies remains the same as in the case with no data contention. DIV outperforms SL in the immediate only case and SL reduces the MDP of class T transactions more than the DIV policy with an increase in the MDP of class NT transactions, in the deferred only and immediate and deferred cases. In gure 14, the SL policy gives a higher MDP than PD at low loads for the T class transactions. This eect is due to the fact that SL policy gives a very high priority to a T class transaction at the beginning. When there is a data conict, it leads to a transaction which has just started to abort a transaction which is in its later stages. The absolute MDP values for these experiments is greater than the MDP values of the previous set of experiments with no data contention, that is, more transactions are aborted due to deadline misses. Due to data conicts more transactions experience waits or get aborted. This shows the eect of data contention on the performance. We are exploring the parameters further to determine if the trend observed also holds for other combination of parameter values, including dierent degrees of data contention, dierent amounts of CPU resources and with disk resident data. 6 Conclusions We have studied the problem of assigning priorities to triggered transactions and reassigning priorities of triggering transactions in a rm real-time active database. We discussed a simple baseline policy PD and two other policies namely DIV and SLACK which assign priorities taking into consideration the active work generated by a transaction. We showed that DIV and SL reduce the deadline miss rate of transactions that trigger other subtransactions (class T) with a small increase in the miss deadline percentage of transactions that don't trigger other transactions (class NT). We showed that DIV outperforms SL when the transactions are triggered only in immediate mode, and that SL favors class T transactions more than DIV when transactions are triggered only in deferred mode. We also showed that SL gives more range exibility to reduce the MDP of T class transactions with an increase in the MDP of NT class transactions, when the transactions are triggered only in 17

18 deferred mode or when they are triggered in immediate or deferred mode with equal probability. We studied the performance of policies in a task only model (no data conicts) and in a main memory database model. The performance of the policies did not dier drastically in these two settings. Some of the extensions we want to address are Experiment with DIV and SL policies on a disk resident database to nd out if disk I/O has an eect on their performance. Experiment with variations of DIV and SLACK policies that have more information about a transaction, for instance the exact number of subtransactions that it is going to trigger, and the type of subtransactions it is going to trigger, i.e., immediate or deferred, to see what advantage is obtained by exploiting this information. Study variations of DIV and SLACK that will assign priorities at every scheduling instance instead of at every object event. Study the eect of errors in the knowledge about transactions like eet and probt rig on the performance of the DIV and SL policies. Explore the relative performance of the DIV and SLACK policies in dierent resource contention regions. References [1] R.K. Abbott, and H. Garcia-Molina, \Scheduling Real-Time Transactions: A Performance Evaluation," ACM Trans. on Database Systems, Sept [2] M.J. Carey, R. Jauhari, and M. Livny, \On Transaction Boundaries in Active Databases: A Performance Perspective," IEEE Trans. on Knowledge and Data Engg., Sept [3] U. Dayal, et. al., \The HIPAC Project: Combining Active Databases and Timing Constraints," SIGMOD Record, 17, 1, March [4] U. Dayal, M. Hsu, and R. Ladin, \Organizing Long-Running Activities with Triggers and Transactions," ACM [5] Special Issue on Active Databases, Data Engineering, Dec [6] J. Huang, J. Stankovic, D. Towsley, and K. Ramamritham, \Experimental Evaluation of Realtime Transaction Processing," IEEE Real-Time Systems Symposium, Dec. 1989, [7] B. Kao, and H. Garcia Molina, \Subtask Deadline Assignment for Complex Distributed Soft Real-Time Tasks," Technical Report STAN-CS , Stanford University, Oct

19 [8] M. H. Klein, T. Ralya, B. Pollak, R. Obenza, and M. G. Harbour, \A Practitioners Handbook for Real-Time Systems - Guide to Rate Monotonic Analysis for Rea-Time Systems," Kluwer Academic Publishers, [9] E. L. Lawler, \Recent Results in Theory of Machine Scheduling". Mathematical Programming : The State of the Art, A.Bachem et. al., Springer-Verlag, New York, [10] M. Livny, \DeNet Users Guide," version 1.5, Dept. Comp. Science, Univ. of Wisconsin, Madison, WI [11] D. McCarthy, and U. Dayal, \The Architecture of an Active Data Base Management System," ACM [12] B. Purimetla, R. M. Sivasankaran and J.Stankovic, \A Study of Distributed Real-Time Active Database Applications," IEEE Workshop on Parallel and Distributed Real-time Systems, April [13] K. Ramamritham, \Real-Time Databases," International Journal of Distributed and Parallel Databases, [14] R. M. Sivasankaran, B. Purimetla, J. Stankovic, and K. Ramamritham, \Network Services Databases - A Distributed Active Real-Time Database (DARTDB) Application," IEEE Workshop on Real-Time Applications, May [15] J. Xu and D. L. Parnas, \Scheduling Processes with Release Times, Deadlines, Precedence and Exclusion Relations," IEEE Transactions on Software Engineering, Vol. 16, No. 3, March

20 100 IMM-NoCC Missed Deadline % T/PD NT/PD T/DIV NT/DIV T/SL NT/SL Load Figure 5: 80 DEF-NoCC Missed Deadline % T/PD NT/PD T/DIV NT/DIV T/SL NT/SL Load Figure 6:

21 80 IMMDEF-NoCC Missed Deadline % T/PD NT/PD T/DIV NT/DIV T/SL NT/SL Load Figure 7: IMM-NoCC::Total TOTAL/PD TOTAL/DIV TOTAL/SL Missed Deadline % Load Figure 8: 21

22 25 DEF-NoCC:Total Missed Deadline % TOTAL/PD TOTAL/DIV TOTAL/SL Load Figure 9: 25 IMMDEF-NoCC::Total Missed Deadline % TOTAL/PD TOTAL/DIV TOTAL/SL Load Figure 10: 22

23 70 65 ALPHA = 0.0 Imm-presc ALPHA-SL ALPHA-DIV Class T Missed Deadline % ALPHA = 1.0 ALPHA = Class NT Missed Deadline % Figure 11: Class T Missed Deadline % ALPHA = 0.0 def-presc ALPHA = 1.0 ALPHA-SL ALPHA-DIV 5 ALPHA = Class NT Missed Deadline % Figure 12: 23

24 60 55 ALPHA = 0.0 Immdef-presc ALPHA-SL ALPHA-DIV Class T Missed Deadline % ALPHA = ALPHA = Class NT Missed Deadline % Figure 13: 70 IMM-CC Missed Deadline % T/PD NT/PD T/DIV NT/DIV T/SL NT/SL Load Figure 14: 24

25 Missed Deadline % T/PD NT/PD T/DIV NT/DIV T/SL NT/SL DEF-CC Load Figure 15: 60 IMMDEF-CC Missed Deadline % T/PD NT/PD T/DIV NT/DIV T/SL NT/SL Load Figure 16: 25

Priority assignment in real-time active databases 1

Priority assignment in real-time active databases 1 The VLDB Journal 1996) 5: 19 34 The VLDB Journal c Springer-Verlag 1996 Priority assignment in real-time active databases 1 Rajendran M. Sivasankaran, John A. Stankovic, Don Towsley, Bhaskar Purimetla,

More information

Performance Evaluation of Two New Disk Scheduling Algorithms. for Real-Time Systems. Department of Computer & Information Science

Performance Evaluation of Two New Disk Scheduling Algorithms. for Real-Time Systems. Department of Computer & Information Science Performance Evaluation of Two New Disk Scheduling Algorithms for Real-Time Systems Shenze Chen James F. Kurose John A. Stankovic Don Towsley Department of Computer & Information Science University of Massachusetts

More information

Maintaining Temporal Consistency: Issues and Algorithms

Maintaining Temporal Consistency: Issues and Algorithms Maintaining Temporal Consistency: Issues and Algorithms Ming Xiong, John A. Stankovic, Krithi Ramamritham, Don Towsley, Rajendran Sivasankaran Department of Computer Science University of Massachusetts

More information

REAL-TIME DATABASES (RTDB)

REAL-TIME DATABASES (RTDB) REAL-TIME DATABASES (RTDB) Prof. Fabio A. Schreiber Politecnico di Milano REAL-TIME DATABASE SYSTEMS DATABASE SYSTEMS WHICH MANAGE TIME CONSTRAINED DATA SPECIFIC TIME INTERVALS FOR DATA VALIDITY VALIDITY

More information

Areal-time database system is a transaction processing

Areal-time database system is a transaction processing IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING, VOL. 14, NO. 5, SEPTEMBER/OCTOBER 2002 1155 Scheduling Transactions with Temporal Constraints: Exploiting Data Semantics Ming Xiong, Member, IEEE, Krithi

More information

Locking Based Concurrency Control. for Integrated Real-Time Database Systems. D. Hong y. S. Chakravarthy. T. Johnson

Locking Based Concurrency Control. for Integrated Real-Time Database Systems. D. Hong y. S. Chakravarthy. T. Johnson Locking Based Concurrency Control for Integrated Real-Time Database Systems D. Hong y S. Chakravarthy T. Johnson Database Systems Research and Development Center Computer and Information Science and Engineering

More information

Research Trends and Issues in Real-Time Database Systems. Sang H. Son. University of Virginia ABSTRACT

Research Trends and Issues in Real-Time Database Systems. Sang H. Son. University of Virginia ABSTRACT Research Trends and Issues in Real-Time Database Systems Sang H. Son Dept. of Computer Science University of Virginia Charlottesville, Virginia 22903 son@cs.virginia.edu ABSTRACT Real-time database systems

More information

Dynamic Voltage Scaling of Periodic and Aperiodic Tasks in Priority-Driven Systems Λ

Dynamic Voltage Scaling of Periodic and Aperiodic Tasks in Priority-Driven Systems Λ Dynamic Voltage Scaling of Periodic and Aperiodic Tasks in Priority-Driven Systems Λ Dongkun Shin Jihong Kim School of CSE School of CSE Seoul National University Seoul National University Seoul, Korea

More information

Real-time Optimistic Concurrency Control based on Transaction Finish Degree

Real-time Optimistic Concurrency Control based on Transaction Finish Degree Journal of Computer Science 1 (4): 471-476, 2005 ISSN 1549-3636 Science Publications, 2005 Real-time Optimistic Concurrency Control based on Transaction Finish Degree 1 Han Qilong, 1,2 Hao Zhongxiao 1

More information

Node Application Logic. SCI Interface. Output FIFO. Input FIFO. Bypass FIFO M U X. Output Link. Input Link. Address Decoder

Node Application Logic. SCI Interface. Output FIFO. Input FIFO. Bypass FIFO M U X. Output Link. Input Link. Address Decoder Real-Time Message Transmission Over The Scalable Coherent Interface (SCI) Lei Jiang Sarit Mukherjee Dept. of Computer Science & Engg. University of Nebraska-Lincoln Lincoln, NE 68588-0115 Email: fljiang,

More information

Real-Time Scalability of Nested Spin Locks. Hiroaki Takada and Ken Sakamura. Faculty of Science, University of Tokyo

Real-Time Scalability of Nested Spin Locks. Hiroaki Takada and Ken Sakamura. Faculty of Science, University of Tokyo Real-Time Scalability of Nested Spin Locks Hiroaki Takada and Ken Sakamura Department of Information Science, Faculty of Science, University of Tokyo 7-3-1, Hongo, Bunkyo-ku, Tokyo 113, Japan Abstract

More information

Effects of Hard Real-Time Constraints in Implementing the Myopic Scheduling Algorithm

Effects of Hard Real-Time Constraints in Implementing the Myopic Scheduling Algorithm Effects of Hard Real-Time Constraints in Implementing the Myopic Scheduling Algorithm Abstract- Institute of Information Technology, University of Dhaka, Dhaka 1 muheymin@yahoo.com, K M. Sakib, M S. Hasan

More information

Issues in Server Farm Design for Real Time E-commerce Transactions

Issues in Server Farm Design for Real Time E-commerce Transactions Issues in Server Farm Design for Real Time E-commerce Transactions Alok Chaturvedi (alok@mgmt.purdue.edu) Krannert Graduate School of Management 3 Krannert Building West Lafayette IN 4796-3 Telephone:

More information

SAMOS: an Active Object{Oriented Database System. Stella Gatziu, Klaus R. Dittrich. Database Technology Research Group

SAMOS: an Active Object{Oriented Database System. Stella Gatziu, Klaus R. Dittrich. Database Technology Research Group SAMOS: an Active Object{Oriented Database System Stella Gatziu, Klaus R. Dittrich Database Technology Research Group Institut fur Informatik, Universitat Zurich fgatziu, dittrichg@ifi.unizh.ch to appear

More information

is developed which describe the mean values of various system parameters. These equations have circular dependencies and must be solved iteratively. T

is developed which describe the mean values of various system parameters. These equations have circular dependencies and must be solved iteratively. T A Mean Value Analysis Multiprocessor Model Incorporating Superscalar Processors and Latency Tolerating Techniques 1 David H. Albonesi Israel Koren Department of Electrical and Computer Engineering University

More information

IBM Almaden Research Center, at regular intervals to deliver smooth playback of video streams. A video-on-demand

IBM Almaden Research Center, at regular intervals to deliver smooth playback of video streams. A video-on-demand 1 SCHEDULING IN MULTIMEDIA SYSTEMS A. L. Narasimha Reddy IBM Almaden Research Center, 650 Harry Road, K56/802, San Jose, CA 95120, USA ABSTRACT In video-on-demand multimedia systems, the data has to be

More information

Number of bits in the period of 100 ms. Number of bits in the period of 100 ms. Number of bits in the periods of 100 ms

Number of bits in the period of 100 ms. Number of bits in the period of 100 ms. Number of bits in the periods of 100 ms Network Bandwidth Reservation using the Rate-Monotonic Model Sourav Ghosh and Ragunathan (Raj) Rajkumar Real-time and Multimedia Systems Laboratory Department of Electrical and Computer Engineering Carnegie

More information

Fault tolerant scheduling in real time systems

Fault tolerant scheduling in real time systems tolerant scheduling in real time systems Afrin Shafiuddin Department of Electrical and Computer Engineering University of Wisconsin-Madison shafiuddin@wisc.edu Swetha Srinivasan Department of Electrical

More information

Scheduling Periodic and Aperiodic. John P. Lehoczky and Sandra R. Thuel. and both hard and soft deadline aperiodic tasks using xed-priority methods.

Scheduling Periodic and Aperiodic. John P. Lehoczky and Sandra R. Thuel. and both hard and soft deadline aperiodic tasks using xed-priority methods. Chapter 8 Scheduling Periodic and Aperiodic Tasks Using the Slack Stealing Algorithm John P. Lehoczky and Sandra R. Thuel This chapter discusses the problem of jointly scheduling hard deadline periodic

More information

Efficient Priority Assignment Policies for Distributed Real-Time Database Systems

Efficient Priority Assignment Policies for Distributed Real-Time Database Systems Proceedings of the 7 WSEAS International Conference on Computer Engineering and Applications, Gold Coast, Australia, January 17-19, 7 7 Efficient Priority Assignment Policies for Distributed Real-Time

More information

MIRROR: A State-Conscious Concurrency Control Protocol for Replicated Real-Time Databases

MIRROR: A State-Conscious Concurrency Control Protocol for Replicated Real-Time Databases : A State-Conscious Concurrency Control Protocol for Replicated Real-Time Databases Ming Xiong y, Krithi Ramamritham z Department of Computer Science University of Massachusetts Amherst, MA 13 fxiong,

More information

ENERGY EFFICIENT SCHEDULING FOR REAL-TIME EMBEDDED SYSTEMS WITH PRECEDENCE AND RESOURCE CONSTRAINTS

ENERGY EFFICIENT SCHEDULING FOR REAL-TIME EMBEDDED SYSTEMS WITH PRECEDENCE AND RESOURCE CONSTRAINTS ENERGY EFFICIENT SCHEDULING FOR REAL-TIME EMBEDDED SYSTEMS WITH PRECEDENCE AND RESOURCE CONSTRAINTS Santhi Baskaran 1 and P. Thambidurai 2 1 Department of Information Technology, Pondicherry Engineering

More information

Concurrent activities in daily life. Real world exposed programs. Scheduling of programs. Tasks in engine system. Engine system

Concurrent activities in daily life. Real world exposed programs. Scheduling of programs. Tasks in engine system. Engine system Real world exposed programs Programs written to interact with the real world, outside the computer Programs handle input and output of data in pace matching the real world processes Necessitates ability

More information

Computer Science 4500 Operating Systems

Computer Science 4500 Operating Systems Computer Science 4500 Operating Systems Module 6 Process Scheduling Methods Updated: September 25, 2014 2008 Stanley A. Wileman, Jr. Operating Systems Slide 1 1 In This Module Batch and interactive workloads

More information

A taxonomy of race. D. P. Helmbold, C. E. McDowell. September 28, University of California, Santa Cruz. Santa Cruz, CA

A taxonomy of race. D. P. Helmbold, C. E. McDowell. September 28, University of California, Santa Cruz. Santa Cruz, CA A taxonomy of race conditions. D. P. Helmbold, C. E. McDowell UCSC-CRL-94-34 September 28, 1994 Board of Studies in Computer and Information Sciences University of California, Santa Cruz Santa Cruz, CA

More information

An Evaluation of Information Retrieval Accuracy. with Simulated OCR Output. K. Taghva z, and J. Borsack z. University of Massachusetts, Amherst

An Evaluation of Information Retrieval Accuracy. with Simulated OCR Output. K. Taghva z, and J. Borsack z. University of Massachusetts, Amherst An Evaluation of Information Retrieval Accuracy with Simulated OCR Output W.B. Croft y, S.M. Harding y, K. Taghva z, and J. Borsack z y Computer Science Department University of Massachusetts, Amherst

More information

MIRROR: A State-Conscious Concurrency Control Protocol for Replicated Real-Time Databases

MIRROR: A State-Conscious Concurrency Control Protocol for Replicated Real-Time Databases University of Massachusetts Amherst ScholarWorks@UMass Amherst Computer Science Department Faculty Publication Series Computer Science 1999 : A State-Conscious Concurrency Control Protocol for Replicated

More information

Multimedia Systems 2011/2012

Multimedia Systems 2011/2012 Multimedia Systems 2011/2012 System Architecture Prof. Dr. Paul Müller University of Kaiserslautern Department of Computer Science Integrated Communication Systems ICSY http://www.icsy.de Sitemap 2 Hardware

More information

requests or displaying activities, hence they usually have soft deadlines, or no deadlines at all. Aperiodic tasks with hard deadlines are called spor

requests or displaying activities, hence they usually have soft deadlines, or no deadlines at all. Aperiodic tasks with hard deadlines are called spor Scheduling Aperiodic Tasks in Dynamic Priority Systems Marco Spuri and Giorgio Buttazzo Scuola Superiore S.Anna, via Carducci 4, 561 Pisa, Italy Email: spuri@fastnet.it, giorgio@sssup.it Abstract In this

More information

On Checkpoint Latency. Nitin H. Vaidya. In the past, a large number of researchers have analyzed. the checkpointing and rollback recovery scheme

On Checkpoint Latency. Nitin H. Vaidya. In the past, a large number of researchers have analyzed. the checkpointing and rollback recovery scheme On Checkpoint Latency Nitin H. Vaidya Department of Computer Science Texas A&M University College Station, TX 77843-3112 E-mail: vaidya@cs.tamu.edu Web: http://www.cs.tamu.edu/faculty/vaidya/ Abstract

More information

Simplified design flow for embedded systems

Simplified design flow for embedded systems Simplified design flow for embedded systems 2005/12/02-1- Reuse of standard software components Knowledge from previous designs to be made available in the form of intellectual property (IP, for SW & HW).

More information

Multiprocessor and Real-Time Scheduling. Chapter 10

Multiprocessor and Real-Time Scheduling. Chapter 10 Multiprocessor and Real-Time Scheduling Chapter 10 1 Roadmap Multiprocessor Scheduling Real-Time Scheduling Linux Scheduling Unix SVR4 Scheduling Windows Scheduling Classifications of Multiprocessor Systems

More information

Using the Imprecise-Computation Technique for Congestion. Control on a Real-Time Trac Switching Element

Using the Imprecise-Computation Technique for Congestion. Control on a Real-Time Trac Switching Element Appeared in Proc. of the Int'l Conf. on Parallel & Distributed Systems, Hsinchu, Taiwan, December 994. Using the Imprecise-Computation Technique for Congestion Control on a Real-Time Trac Switching Element

More information

Mixed Criticality Scheduling in Time-Triggered Legacy Systems

Mixed Criticality Scheduling in Time-Triggered Legacy Systems Mixed Criticality Scheduling in Time-Triggered Legacy Systems Jens Theis and Gerhard Fohler Technische Universität Kaiserslautern, Germany Email: {jtheis,fohler}@eit.uni-kl.de Abstract Research on mixed

More information

Introduction to Embedded Systems

Introduction to Embedded Systems Introduction to Embedded Systems Sanjit A. Seshia UC Berkeley EECS 9/9A Fall 0 008-0: E. A. Lee, A. L. Sangiovanni-Vincentelli, S. A. Seshia. All rights reserved. Chapter : Operating Systems, Microkernels,

More information

Towards the Evaluation of Algorithms used in Real-Time Databases

Towards the Evaluation of Algorithms used in Real-Time Databases Towards the Evaluation of Algorithms used in Real-Time Databases VÁCLAV KRÓL 1, JINDŘICH ČERNOHORSKÝ 2, JAN POKORNÝ 2 1 Institute of Information Technologies Silesian University in Opava, Faculty of Business

More information

Real-Time Scheduling. Dynamic Priority Servers

Real-Time Scheduling. Dynamic Priority Servers Real-Time Scheduling Dynamic Priority Servers Objectives Schedule soft aperiodic and hard periodic tasks Reduce average response time of aperiodic requests without compromising schedulability of periodic

More information

REAL-TIME SCHEDULING OF SOFT PERIODIC TASKS ON MULTIPROCESSOR SYSTEMS: A FUZZY MODEL

REAL-TIME SCHEDULING OF SOFT PERIODIC TASKS ON MULTIPROCESSOR SYSTEMS: A FUZZY MODEL Available Online at www.ijcsmc.com International Journal of Computer Science and Mobile Computing A Monthly Journal of Computer Science and Information Technology IJCSMC, Vol. 3, Issue. 6, June 2014, pg.348

More information

Consistent Logical Checkpointing. Nitin H. Vaidya. Texas A&M University. Phone: Fax:

Consistent Logical Checkpointing. Nitin H. Vaidya. Texas A&M University. Phone: Fax: Consistent Logical Checkpointing Nitin H. Vaidya Department of Computer Science Texas A&M University College Station, TX 77843-3112 hone: 409-845-0512 Fax: 409-847-8578 E-mail: vaidya@cs.tamu.edu Technical

More information

Processing Real-Time Transactions in a Replicated Database System

Processing Real-Time Transactions in a Replicated Database System Distributed and Parallel Databases, 2, 405436 (1994) 1994 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands. Processing Real-Time Transactions in a Replicated Database System C)ZGOR ULUSOY

More information

SCHEDULING REAL-TIME MESSAGES IN PACKET-SWITCHED NETWORKS IAN RAMSAY PHILP. B.S., University of North Carolina at Chapel Hill, 1988

SCHEDULING REAL-TIME MESSAGES IN PACKET-SWITCHED NETWORKS IAN RAMSAY PHILP. B.S., University of North Carolina at Chapel Hill, 1988 SCHEDULING REAL-TIME MESSAGES IN PACKET-SWITCHED NETWORKS BY IAN RAMSAY PHILP B.S., University of North Carolina at Chapel Hill, 1988 M.S., University of Florida, 1990 THESIS Submitted in partial fulllment

More information

A Study of the Performance Tradeoffs of a Tape Archive

A Study of the Performance Tradeoffs of a Tape Archive A Study of the Performance Tradeoffs of a Tape Archive Jason Xie (jasonxie@cs.wisc.edu) Naveen Prakash (naveen@cs.wisc.edu) Vishal Kathuria (vishal@cs.wisc.edu) Computer Sciences Department University

More information

(Preliminary Version 2 ) Jai-Hoon Kim Nitin H. Vaidya. Department of Computer Science. Texas A&M University. College Station, TX

(Preliminary Version 2 ) Jai-Hoon Kim Nitin H. Vaidya. Department of Computer Science. Texas A&M University. College Station, TX Towards an Adaptive Distributed Shared Memory (Preliminary Version ) Jai-Hoon Kim Nitin H. Vaidya Department of Computer Science Texas A&M University College Station, TX 77843-3 E-mail: fjhkim,vaidyag@cs.tamu.edu

More information

Performance of Multihop Communications Using Logical Topologies on Optical Torus Networks

Performance of Multihop Communications Using Logical Topologies on Optical Torus Networks Performance of Multihop Communications Using Logical Topologies on Optical Torus Networks X. Yuan, R. Melhem and R. Gupta Department of Computer Science University of Pittsburgh Pittsburgh, PA 156 fxyuan,

More information

Analytic Performance Models for Bounded Queueing Systems

Analytic Performance Models for Bounded Queueing Systems Analytic Performance Models for Bounded Queueing Systems Praveen Krishnamurthy Roger D. Chamberlain Praveen Krishnamurthy and Roger D. Chamberlain, Analytic Performance Models for Bounded Queueing Systems,

More information

Implementing Scheduling Algorithms. Real-Time and Embedded Systems (M) Lecture 9

Implementing Scheduling Algorithms. Real-Time and Embedded Systems (M) Lecture 9 Implementing Scheduling Algorithms Real-Time and Embedded Systems (M) Lecture 9 Lecture Outline Implementing real time systems Key concepts and constraints System architectures: Cyclic executive Microkernel

More information

Disk Scheduling with Dynamic Request Priorities. Yongcheng Li See-Mong Tan Zhigang Chen Roy H. Campbell. Department of Computer Science

Disk Scheduling with Dynamic Request Priorities. Yongcheng Li See-Mong Tan Zhigang Chen Roy H. Campbell. Department of Computer Science Disk Scheduling with Dynamic Request Priorities Yongcheng Li See-Mong Tan Zhigang Chen Roy H. Campbell Department of Computer Science University of Illinois at Urbana-Champaign Digital Computer Laboratory

More information

ENERGY EFFICIENT SCHEDULING SIMULATOR FOR DISTRIBUTED REAL-TIME SYSTEMS

ENERGY EFFICIENT SCHEDULING SIMULATOR FOR DISTRIBUTED REAL-TIME SYSTEMS I J I T E ISSN: 2229-7367 3(1-2), 2012, pp. 409-414 ENERGY EFFICIENT SCHEDULING SIMULATOR FOR DISTRIBUTED REAL-TIME SYSTEMS SANTHI BASKARAN 1, VARUN KUMAR P. 2, VEVAKE B. 2 & KARTHIKEYAN A. 2 1 Assistant

More information

Allowing Cycle-Stealing Direct Memory Access I/O. Concurrent with Hard-Real-Time Programs

Allowing Cycle-Stealing Direct Memory Access I/O. Concurrent with Hard-Real-Time Programs To appear in: Int. Conf. on Parallel and Distributed Systems, ICPADS'96, June 3-6, 1996, Tokyo Allowing Cycle-Stealing Direct Memory Access I/O Concurrent with Hard-Real-Time Programs Tai-Yi Huang, Jane

More information

ECE519 Advanced Operating Systems

ECE519 Advanced Operating Systems IT 540 Operating Systems ECE519 Advanced Operating Systems Prof. Dr. Hasan Hüseyin BALIK (10 th Week) (Advanced) Operating Systems 10. Multiprocessor, Multicore and Real-Time Scheduling 10. Outline Multiprocessor

More information

RTC: Language Support for Real-Time Concurrency

RTC: Language Support for Real-Time Concurrency RTC: Language Support for Real-Time Concurrency Insup Lee, Susan Davidson, and Victor Wolfe 1 Introduction The RTC (Real-Time Concurrency) programming concepts and language constructs for expressing timing

More information

Multimedia Applications Require Adaptive CPU Scheduling. Veronica Baiceanu, Crispin Cowan, Dylan McNamee, Calton Pu, and Jonathan Walpole

Multimedia Applications Require Adaptive CPU Scheduling. Veronica Baiceanu, Crispin Cowan, Dylan McNamee, Calton Pu, and Jonathan Walpole Multimedia Applications Require Adaptive CPU Scheduling Veronica Baiceanu, Crispin Cowan, Dylan McNamee, Calton Pu, and Jonathan Walpole Department of Computer Science and Engineering Oregon Graduate Institute

More information

Relative Reduced Hops

Relative Reduced Hops GreedyDual-Size: A Cost-Aware WWW Proxy Caching Algorithm Pei Cao Sandy Irani y 1 Introduction As the World Wide Web has grown in popularity in recent years, the percentage of network trac due to HTTP

More information

Optimistic Concurrency Control Methods for Real-Time Database Systems

Optimistic Concurrency Control Methods for Real-Time Database Systems DEPARTMENT OF COMPUTER SCIENCE SERIES OF PUBLICATIONS C REPORT C-2001-9 Optimistic Concurrency Control Methods for Real-Time Database Systems Jan Lindström UNIVERSITY OF HELSINKI FINLAND Optimistic Concurrency

More information

Precedence Graphs Revisited (Again)

Precedence Graphs Revisited (Again) Precedence Graphs Revisited (Again) [i,i+6) [i+6,i+12) T 2 [i,i+6) [i+6,i+12) T 3 [i,i+2) [i+2,i+4) [i+4,i+6) [i+6,i+8) T 4 [i,i+1) [i+1,i+2) [i+2,i+3) [i+3,i+4) [i+4,i+5) [i+5,i+6) [i+6,i+7) T 5 [i,i+1)

More information

CS4514 Real Time Scheduling

CS4514 Real Time Scheduling CS4514 Real Time Scheduling Jose M. Garrido Fall 2015 Department of Computer Science 1 Periodic Tasks Typical real-time application has many tasks that need to be executed periodically Reading sensor data

More information

MDP Routing in ATM Networks. Using the Virtual Path Concept 1. Department of Computer Science Department of Computer Science

MDP Routing in ATM Networks. Using the Virtual Path Concept 1. Department of Computer Science Department of Computer Science MDP Routing in ATM Networks Using the Virtual Path Concept 1 Ren-Hung Hwang, James F. Kurose, and Don Towsley Department of Computer Science Department of Computer Science & Information Engineering University

More information

Predicting the Worst-Case Execution Time of the Concurrent Execution. of Instructions and Cycle-Stealing DMA I/O Operations

Predicting the Worst-Case Execution Time of the Concurrent Execution. of Instructions and Cycle-Stealing DMA I/O Operations ACM SIGPLAN Workshop on Languages, Compilers and Tools for Real-Time Systems, La Jolla, California, June 1995. Predicting the Worst-Case Execution Time of the Concurrent Execution of Instructions and Cycle-Stealing

More information

Scheduling in Multiprocessor System Using Genetic Algorithms

Scheduling in Multiprocessor System Using Genetic Algorithms Scheduling in Multiprocessor System Using Genetic Algorithms Keshav Dahal 1, Alamgir Hossain 1, Benzy Varghese 1, Ajith Abraham 2, Fatos Xhafa 3, Atanasi Daradoumis 4 1 University of Bradford, UK, {k.p.dahal;

More information

Two-Version-Based Concurrency Control and Recovery in Real-Time Client/Server Databases

Two-Version-Based Concurrency Control and Recovery in Real-Time Client/Server Databases Two-Version-Based Concurrency Control and Recovery in Real-Time Client/Server Databases Tei-Wei Kuo, Yuan-Ting Kao, and Chin-Fu Kuo Department of Computer Science and Information Engineering National Taiwan

More information

Ecient Processor Allocation for 3D Tori. Wenjian Qiao and Lionel M. Ni. Department of Computer Science. Michigan State University

Ecient Processor Allocation for 3D Tori. Wenjian Qiao and Lionel M. Ni. Department of Computer Science. Michigan State University Ecient Processor llocation for D ori Wenjian Qiao and Lionel M. Ni Department of Computer Science Michigan State University East Lansing, MI 4884-107 fqiaow, nig@cps.msu.edu bstract Ecient allocation of

More information

Recovering from Main-Memory Lapses. H.V. Jagadish Avi Silberschatz S. Sudarshan. AT&T Bell Labs. 600 Mountain Ave., Murray Hill, NJ 07974

Recovering from Main-Memory Lapses. H.V. Jagadish Avi Silberschatz S. Sudarshan. AT&T Bell Labs. 600 Mountain Ave., Murray Hill, NJ 07974 Recovering from Main-Memory Lapses H.V. Jagadish Avi Silberschatz S. Sudarshan AT&T Bell Labs. 600 Mountain Ave., Murray Hill, NJ 07974 fjag,silber,sudarshag@allegra.att.com Abstract Recovery activities,

More information

Memory hierarchy. 1. Module structure. 2. Basic cache memory. J. Daniel García Sánchez (coordinator) David Expósito Singh Javier García Blas

Memory hierarchy. 1. Module structure. 2. Basic cache memory. J. Daniel García Sánchez (coordinator) David Expósito Singh Javier García Blas Memory hierarchy J. Daniel García Sánchez (coordinator) David Expósito Singh Javier García Blas Computer Architecture ARCOS Group Computer Science and Engineering Department University Carlos III of Madrid

More information

Implementation and modeling of two-phase locking concurrency control a performance study

Implementation and modeling of two-phase locking concurrency control a performance study INFSOF 4047 Information and Software Technology 42 (2000) 257 273 www.elsevier.nl/locate/infsof Implementation and modeling of two-phase locking concurrency control a performance study N.B. Al-Jumah a,

More information

A Boolean Expression. Reachability Analysis or Bisimulation. Equation Solver. Boolean. equations.

A Boolean Expression. Reachability Analysis or Bisimulation. Equation Solver. Boolean. equations. A Framework for Embedded Real-time System Design? Jin-Young Choi 1, Hee-Hwan Kwak 2, and Insup Lee 2 1 Department of Computer Science and Engineering, Korea Univerity choi@formal.korea.ac.kr 2 Department

More information

Performance Comparison Between AAL1, AAL2 and AAL5

Performance Comparison Between AAL1, AAL2 and AAL5 The University of Kansas Technical Report Performance Comparison Between AAL1, AAL2 and AAL5 Raghushankar R. Vatte and David W. Petr ITTC-FY1998-TR-13110-03 March 1998 Project Sponsor: Sprint Corporation

More information

Steering. Stream. User Interface. Stream. Manager. Interaction Managers. Snapshot. Stream

Steering. Stream. User Interface. Stream. Manager. Interaction Managers. Snapshot. Stream Agent Roles in Snapshot Assembly Delbert Hart Dept. of Computer Science Washington University in St. Louis St. Louis, MO 63130 hart@cs.wustl.edu Eileen Kraemer Dept. of Computer Science University of Georgia

More information

Networks for Control. California Institute of Technology. Pasadena, CA Abstract

Networks for Control. California Institute of Technology. Pasadena, CA Abstract Learning Fuzzy Rule-Based Neural Networks for Control Charles M. Higgins and Rodney M. Goodman Department of Electrical Engineering, 116-81 California Institute of Technology Pasadena, CA 91125 Abstract

More information

PROCESS SCHEDULING II. CS124 Operating Systems Fall , Lecture 13

PROCESS SCHEDULING II. CS124 Operating Systems Fall , Lecture 13 PROCESS SCHEDULING II CS124 Operating Systems Fall 2017-2018, Lecture 13 2 Real-Time Systems Increasingly common to have systems with real-time scheduling requirements Real-time systems are driven by specific

More information

Real-time operating systems and scheduling

Real-time operating systems and scheduling Real-time operating systems and scheduling Problem 21 Consider a real-time operating system (OS) that has a built-in preemptive scheduler. Each task has a unique priority and the lower the priority id,

More information

Comparing two-phase locking and optimistic concurrency control protocols in multiprocessor real-time databases

Comparing two-phase locking and optimistic concurrency control protocols in multiprocessor real-time databases Title Comparing two-phase locking and optimistic concurrency control protocols in multiprocessor real-time databases Author(s) Chiu, A; Kao, CM; Lam, KY Citation The 5th Joint Workshop on Parallel and

More information

In Proceedings of ICDCS 97: The IEEE International Conference on Distributed Computing Systems, Baltimore, Maryland, May 1997.

In Proceedings of ICDCS 97: The IEEE International Conference on Distributed Computing Systems, Baltimore, Maryland, May 1997. In Proceedings of ICDCS 97: The IEEE International Conference on Distributed Computing Systems, Baltimore, Maryland, May 1997. Load Profiling A Methodology for Scheduling Real-Time Tasks in a Distributed

More information

Submitted for TAU97 Abstract Many attempts have been made to combine some form of retiming with combinational

Submitted for TAU97 Abstract Many attempts have been made to combine some form of retiming with combinational Experiments in the Iterative Application of Resynthesis and Retiming Soha Hassoun and Carl Ebeling Department of Computer Science and Engineering University ofwashington, Seattle, WA fsoha,ebelingg@cs.washington.edu

More information

On Checkpoint Latency. Nitin H. Vaidya. Texas A&M University. Phone: (409) Technical Report

On Checkpoint Latency. Nitin H. Vaidya. Texas A&M University.   Phone: (409) Technical Report On Checkpoint Latency Nitin H. Vaidya Department of Computer Science Texas A&M University College Station, TX 77843-3112 E-mail: vaidya@cs.tamu.edu Phone: (409) 845-0512 FAX: (409) 847-8578 Technical Report

More information

A Transaction Processing Technique in Real-Time Object- Oriented Databases

A Transaction Processing Technique in Real-Time Object- Oriented Databases 122 IJCSNS International Journal of Computer Science and Network Security, VOL.8 No.1, January 2008 A Transaction Processing Technique in Real-Time Object- Oriented Databases Woochun Jun Dept. of Computer

More information

In Proceedings of RTAS 97: The 1997 IEEE Real-Time Technology and Applications Symposium, Montreal, Canada, June 1997.

In Proceedings of RTAS 97: The 1997 IEEE Real-Time Technology and Applications Symposium, Montreal, Canada, June 1997. In Proceedings of RTAS 97: The 1997 IEEE Real-Time Technology and Applications Symposium, Montreal, Canada, June 1997. Admission Control for Soft-Deadline Transactions in ACCORD Sue Nagy y Azer Bestavros

More information

The Implicit{Yes Vote Commit Protocol. with Delegation of Commitment. Pittsburgh, PA Pittsburgh, PA are relatively less reliable.

The Implicit{Yes Vote Commit Protocol. with Delegation of Commitment. Pittsburgh, PA Pittsburgh, PA are relatively less reliable. The Implicit{Yes Vote Commit Protocol with Delegation of Commitment Yousef J. Al-Houmaily Panos K. Chrysanthis Dept. of Electrical Engineering Dept. of Computer Science University of Pittsburgh University

More information

Analytical Modeling of Routing Algorithms in. Virtual Cut-Through Networks. Real-Time Computing Laboratory. Electrical Engineering & Computer Science

Analytical Modeling of Routing Algorithms in. Virtual Cut-Through Networks. Real-Time Computing Laboratory. Electrical Engineering & Computer Science Analytical Modeling of Routing Algorithms in Virtual Cut-Through Networks Jennifer Rexford Network Mathematics Research Networking & Distributed Systems AT&T Labs Research Florham Park, NJ 07932 jrex@research.att.com

More information

A Concurrency Control for Transactional Mobile Agents

A Concurrency Control for Transactional Mobile Agents A Concurrency Control for Transactional Mobile Agents Jeong-Joon Yoo and Dong-Ik Lee Department of Information and Communications, Kwang-Ju Institute of Science and Technology (K-JIST) Puk-Gu Oryong-Dong

More information

highest cosine coecient [5] are returned. Notice that a query can hit documents without having common terms because the k indexing dimensions indicate

highest cosine coecient [5] are returned. Notice that a query can hit documents without having common terms because the k indexing dimensions indicate Searching Information Servers Based on Customized Proles Technical Report USC-CS-96-636 Shih-Hao Li and Peter B. Danzig Computer Science Department University of Southern California Los Angeles, California

More information

On Hierarchical Scheduling in VxWorks

On Hierarchical Scheduling in VxWorks On Hierarchical Scheduling in VxWorks by Mikael Åsberg (mag04002@student.mdh.se) Master thesis Supervisor: Moris Behnam Examiner: Thomas Nolte Mälardalen University Department of Computer Science and Electronics

More information

Incompatibility Dimensions and Integration of Atomic Commit Protocols

Incompatibility Dimensions and Integration of Atomic Commit Protocols The International Arab Journal of Information Technology, Vol. 5, No. 4, October 2008 381 Incompatibility Dimensions and Integration of Atomic Commit Protocols Yousef Al-Houmaily Department of Computer

More information

Synchronization Part II. CS403/534 Distributed Systems Erkay Savas Sabanci University

Synchronization Part II. CS403/534 Distributed Systems Erkay Savas Sabanci University Synchronization Part II CS403/534 Distributed Systems Erkay Savas Sabanci University 1 Election Algorithms Issue: Many distributed algorithms require that one process act as a coordinator (initiator, etc).

More information

Determining the Number of CPUs for Query Processing

Determining the Number of CPUs for Query Processing Determining the Number of CPUs for Query Processing Fatemah Panahi Elizabeth Soechting CS747 Advanced Computer Systems Analysis Techniques The University of Wisconsin-Madison fatemeh@cs.wisc.edu, eas@cs.wisc.edu

More information

An Approach to Task Attribute Assignment for Uniprocessor Systems

An Approach to Task Attribute Assignment for Uniprocessor Systems An Approach to ttribute Assignment for Uniprocessor Systems I. Bate and A. Burns Real-Time Systems Research Group Department of Computer Science University of York York, United Kingdom e-mail: fijb,burnsg@cs.york.ac.uk

More information

arxiv: v2 [cs.ds] 22 Jun 2016

arxiv: v2 [cs.ds] 22 Jun 2016 Federated Scheduling Admits No Constant Speedup Factors for Constrained-Deadline DAG Task Systems Jian-Jia Chen Department of Informatics, TU Dortmund University, Germany arxiv:1510.07254v2 [cs.ds] 22

More information

Multiprocessor and Real- Time Scheduling. Chapter 10

Multiprocessor and Real- Time Scheduling. Chapter 10 Multiprocessor and Real- Time Scheduling Chapter 10 Classifications of Multiprocessor Loosely coupled multiprocessor each processor has its own memory and I/O channels Functionally specialized processors

More information

On Object Orientation as a Paradigm for General Purpose. Distributed Operating Systems

On Object Orientation as a Paradigm for General Purpose. Distributed Operating Systems On Object Orientation as a Paradigm for General Purpose Distributed Operating Systems Vinny Cahill, Sean Baker, Brendan Tangney, Chris Horn and Neville Harris Distributed Systems Group, Dept. of Computer

More information

Design and Evaluation of a Feedback Control EDF Scheduling Algorithm

Design and Evaluation of a Feedback Control EDF Scheduling Algorithm Clint Moulds cmould1@gl.umbc.edu Design and Evaluation of a Feedback Control EDF Scheduling Algorithm by Lu, Stankovic, Tao, and Son A summary presentation CMSC 691S Younis Spring 2005 Table of Section

More information

On the Relationship of Server Disk Workloads and Client File Requests

On the Relationship of Server Disk Workloads and Client File Requests On the Relationship of Server Workloads and Client File Requests John R. Heath Department of Computer Science University of Southern Maine Portland, Maine 43 Stephen A.R. Houser University Computing Technologies

More information

4/6/2011. Informally, scheduling is. Informally, scheduling is. More precisely, Periodic and Aperiodic. Periodic Task. Periodic Task (Contd.

4/6/2011. Informally, scheduling is. Informally, scheduling is. More precisely, Periodic and Aperiodic. Periodic Task. Periodic Task (Contd. So far in CS4271 Functionality analysis Modeling, Model Checking Timing Analysis Software level WCET analysis System level Scheduling methods Today! erformance Validation Systems CS 4271 Lecture 10 Abhik

More information

instruction fetch memory interface signal unit priority manager instruction decode stack register sets address PC2 PC3 PC4 instructions extern signals

instruction fetch memory interface signal unit priority manager instruction decode stack register sets address PC2 PC3 PC4 instructions extern signals Performance Evaluations of a Multithreaded Java Microcontroller J. Kreuzinger, M. Pfeer A. Schulz, Th. Ungerer Institute for Computer Design and Fault Tolerance University of Karlsruhe, Germany U. Brinkschulte,

More information

MIRROR: A State-Conscious Concurrency Control Protocol for Replicated Real-Time Databases

MIRROR: A State-Conscious Concurrency Control Protocol for Replicated Real-Time Databases : A State-Conscious Concurrency Control Protocol for Replicated Real-Time Databases Ming Xiong æ, Krithi Ramamritham y, Jayant Haritsa z, John A. Stankovic x Abstract Data replication can help database

More information

Performance Evaluations and Estimations of Workload of On-Demand Updates in Soft Real-Time Systems

Performance Evaluations and Estimations of Workload of On-Demand Updates in Soft Real-Time Systems Performance Evaluations and Estimations of Workload of On-Demand Updates in Soft Real-Time Systems Thomas Gustafsson a,b and Jörgen Hansson b,c a School of Engineering, Jönköping University, Sweden b Department

More information

Chapter 16: Distributed Synchronization

Chapter 16: Distributed Synchronization Chapter 16: Distributed Synchronization Chapter 16 Distributed Synchronization Event Ordering Mutual Exclusion Atomicity Concurrency Control Deadlock Handling Election Algorithms Reaching Agreement 18.2

More information

In: Won Kim, editor, Modern Database Systems: The Object Model, Interoperability, and Beyond, Addison-Wesley, Reading, Massachusetts, Sep.

In: Won Kim, editor, Modern Database Systems: The Object Model, Interoperability, and Beyond, Addison-Wesley, Reading, Massachusetts, Sep. In: Won Kim, editor, Modern Database Systems: The Object Model, Interoperability, and Beyond, Addison-Wesley, Reading, Massachusetts, Sep. 1994 Active Database Systems Umeshwar Dayal Hewlett-Packard Laboratories

More information

Overview. Sporadic tasks. Recall. Aperiodic tasks. Real-time Systems D0003E 2/26/2009. Loosening D = T. Aperiodic tasks. Response-time analysis

Overview. Sporadic tasks. Recall. Aperiodic tasks. Real-time Systems D0003E 2/26/2009. Loosening D = T. Aperiodic tasks. Response-time analysis Overview Real-time Systems D0003E Lecture 11: Priority inversion Burns/Wellings ch. 13 (except 13.12) Aperiodic tasks Response time analysis Blocking Priority inversion Priority inheritance Priority ceiling

More information

2 IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, VOL. XX, NO. Y, MONTH 1999 B. Contributions For the above real-time context, the main contrib

2 IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, VOL. XX, NO. Y, MONTH 1999 B. Contributions For the above real-time context, the main contrib IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, VOL. XX, NO. Y, MONTH 1999 1 The PROMPT Real-Time Commit Protocol Jayant R. Haritsa, Krithi Ramamritham, Ramesh Gupta Abstract Weinvestigate the performance

More information

Time Triggered and Event Triggered; Off-line Scheduling

Time Triggered and Event Triggered; Off-line Scheduling Time Triggered and Event Triggered; Off-line Scheduling Real-Time Architectures -TUe Gerhard Fohler 2004 Mälardalen University, Sweden gerhard.fohler@mdh.se Real-time: TT and ET Gerhard Fohler 2004 1 Activation

More information

Skill. Robot/ Controller

Skill. Robot/ Controller Skill Acquisition from Human Demonstration Using a Hidden Markov Model G. E. Hovland, P. Sikka and B. J. McCarragher Department of Engineering Faculty of Engineering and Information Technology The Australian

More information